Krylow A M, Rymer W Z
Department of Biomedical Engineering, Northwestern University, Evanston, IL 60208 USA.
IEEE Trans Biomed Eng. 1997 Feb;44(2):165-76. doi: 10.1109/10.552246.
Human upper limb movement trajectories have been shown to be quite smooth, in that time derivatives of end point position (r), including d3r/dt3 (i.e., jerk), appear to be minimized during rapid voluntary reaching tasks. Studies have suggested that these movements are implemented by an optimal neural controller which seeks to minimize a cost function, such as average jerk cost, over the course of these motions. While this hypothetical control strategy is widely supported, there are substantial difficulties associated with implementing such a controller, including ambiguities inherent in transformations from Cartesian to joint coordinates, and the lack of appropriate transducers to provide information about higher derivatives of limb motion to the nervous system. Given these limitations, we evaluate the possibility that smoothing of movement might be induced primarily by the intrinsic mechanical properties of muscle by recording the trajectories of inertially loaded muscle with the excitatory input held constant. These trajectories are compared with those predicted by a minimum-jerk optimization model, and by a Hill-based muscle model. Our results indicate that trajectories produced by inertially loaded muscle alone are smooth (in the minimum-jerk sense), and that muscle properties may suffice to account for much of the observed smoothing of voluntary motion, obviating the need for an optimizing neural strategy.
研究表明，人类上肢运动轨迹相当平滑，即在快速自主伸手任务中，终点位置（r）的时间导数，包括d3r/dt3（即加加速度）似乎被最小化。研究表明，这些运动是由一个最优神经控制器实现的，该控制器试图在这些运动过程中最小化一个成本函数，如平均加加速度成本。虽然这种假设的控制策略得到了广泛支持，但实施这样一个控制器存在重大困难，包括从笛卡尔坐标到关节坐标转换中固有的模糊性，以及缺乏适当的传感器来向神经系统提供有关肢体运动高阶导数的信息。鉴于这些局限性，我们通过在兴奋性输入保持恒定的情况下记录惯性加载肌肉的轨迹，来评估运动平滑可能主要由肌肉的内在机械特性引起的可能性。将这些轨迹与最小加加速度优化模型和基于希尔的肌肉模型预测的轨迹进行比较。我们的结果表明，仅由惯性加载肌肉产生的轨迹是平滑的（在最小加加速度意义上），并且肌肉特性可能足以解释观察到的大部分自主运动平滑现象，从而无需优化神经策略。
